Most on-chain metrics are useless. That's the blunt assessment from Charles Edwards, founder of Capriole Investments, who argues that 90 to 99 percent of what passes for blockchain data is noise. His critique lands as institutions begin cross-checking crypto dashboards against traditional financial data — and finding serious discrepancies.
The case against on-chain noise
Edwards said his firm builds models from a small, tested set of signals. One example is the Macro Index, which blends over 60 metrics into a single reading. The rest, he argues, just clutters the picture. "Up to 99% of on-chain metrics add nothing but noise," Edwards told GFdaily. His view is shared by Julio Moreno, Head of Research at CryptoQuant, who noted that institutions now verify on-chain data against traditional data before acting on it.
The FTX test case
The danger of relying on raw numbers became clear during the FTX collapse. On November 6, 2022, CryptoQuant's exchange reserves data showed FTX holding 20,177 BTC. By November 8, that number had dropped to 0.64 BTC — a near-total depletion. But other dashboards still showed FTX holding reserves at the same time. The difference came down to methodology: some platforms linked addresses more aggressively, while CryptoQuant stayed conservative. Same metric name, wildly different outputs.
Same metric name, different calculation. That's the norm, not the exception. Data sources vary. Normalization differs. Gaps in time series get handled differently. The result: two dashboards can show contradictory numbers for the same exchange, the same day.
What investors should check
The takeaway isn't that on-chain data is worthless — it's that the method behind the number matters more than the number itself. Edwards and Moreno both urge investors to study how a metric is built before trusting it. Transparency around methodology, they argue, is becoming a competitive edge for analytics firms. The ones that explain their address linking, their gap handling, and their normalization steps will earn more trust. The ones that just flash a big number won't.
